years, with the Compustat data showing the greater variance but a lower
mean value. Thus, the petroleum industry reported higher mean R&D ex-
penditures under the more restrictive NSF/Census definitions than under the
broader SEC definitions.^7 Other than for the petroleum group, the industry-
level data are well behaved. Industry groupings with fewer than 10 cases
were not analyzed separately because of sample size. (There were only
seven cases in the next largest industry after petroleum.)
In summary, the R&D expenditure data for the complete sample are
comparable for all years except 1978 and 1979. When specific industry
data are examined, the petroleum industry data have dissimilar variances
in 1975–1979, and the mean values are uniformly different throughout.
Moreover, the differences in the petroleum industry mean R&D expendi-
tures are in the opposite direction from the complete sample differences in
all years except 1979 and 1982. This suggests that the industry consistently
underreported its R&D expenditures on SEC Forms 10-K and 10-Q, while
the firms in the complete sample tended to overreport, relative to the
NSF/Census figures.
Comparison of Regression Results
The question addressed by the regression results is whether the support for
the imperfect markets hypothesis is strengthened or weakened by substitut-
188 COMPARING CENSUS/NSF R&D DATA WITH COMPUSTAT R&D DATA
TABLE 7.2 Comparison of R&D Expenditure Data for Matched Samples of
Firms: Compustat versus NSF/Census Data, 1975–1982. Z Scores for
Differences in Means*
1975 1976 1977 1978 1979 1980 1981 1982Complete Sample F= 0.004 0.014 0.315 0.308 0.858 0.395 0.200 0.266
N= (188) (188) (188) (188) (180) (177) (162) (158)
Industries
Construction F= 0.222 0.816 0.097 0.029 0.328 0.531 0.360 0.643
N= (16) (16) (16) (16) (16) (16) (17) (17)
Petroleum F= 1.549 –1.229 –1.348 –1.382 –1.148 –1.344 –1.161 –1.500
N= (12) (12) (12) (12) (12) (11) (11) (11)
Machinery F= 0.063 –0.031 –0.018 –0.362 –0.225 –0.490 –0.085 –0.236
N= (16) (16) (16) (16) (16) (16) (15) (15)
Electronics F= 0.003 –0.134 0.364 0.578 0.424 0.375 0.150 0.443
N= (21) (21) (21) (21) (22) (22) (17) (15)
Drugs F= 0.092 –0.027 0.192 0.377 0.319 0.460 0.383 0.313
N= (25) (25) (25) (25) (25) (24) (23) (23)
Chemicals F= 1.076 0.825 1.230 1.008 0.616 0.752 0.793 0.063
N= (24) (24) (24) (24) (24) (24) (23) (23)
*A positive score indicates that the Compustat mean exceeds the NSF/Census mean; a negative score indicates that the
NSF/Census mean exceeds the Compustat mean.